Retrieval Augmented Generation (RAG) allows a researcher to constrain a large language model to use specific source files, essentially providing it with a curated knowledge base to use when answering questions. This allows the service to answer questions based on the knowledge contained in the source files, and to cite its source for its answer.

Objectives: You will get an overview of the affordances of RAG and the processes necessary to build a RAG service based on your own text sources. You will also see demonstrations of RAG services based on some small-scale text collections, with introductions to the tools used to build and run them.

Completion of this workshop can contribute to a partial requirement of the University of Alberta Professional Development program by the Faculty of Graduate & Postdoctoral Studies (GPS). Information on how attendance will be recorded will be discussed in the workshop.

About the Instructor: Peter Binkley is the Digital Scholarship Technologies Librarian, and has been a librarian at the U of A for 24 years. His focus is on helping students and researchers apply digital techniques to their scholarship (broadly defined). This includes areas like the sharing of research online using durable static websites, the use of core digital tools like git and python, and most recently the use of local Large Language Models to apply AI techniques to research problems safely and securely. He has an MLIS and a Ph.D. (in Medieval Studies, working on 13th- and 14th-century Latin texts and manuscripts). Any questions about the workshop can be directed to peter.binkley@ualberta.ca.